#Make a hugging face version of audiosr
1 messages · Page 1 of 1 (latest)
good luck
try not to get as many errors as possible
tip: make the huggingface space gpu friendly
gl @south galleon uwu
Nope, I'm working on another project that will be released soon.
I found this colab, could be useful meanwhile:
https://colab.research.google.com/github/jarredou/AudioSR-Colab-Fork/blob/main/AudioSR_Colab_Fork.ipynb
i already made that https://huggingface.co/spaces/Nick088/Audio-SR
only problem is that its on cpu
and tbh the only thing i dont like of audio sr is that it slowed down the results for me
https://github.com/haoheliu/versatile_audio_super_resolution/issues/52#issuecomment-2161542158 but that seems like an issue of the gradio ui ig? idk really its an abandoned project
btw did u ever have issue of audio sr results being slowed down
Nope
on the gradio version or cli one
CLI
they were slightly slowed down for me
i always used only the gradio one
bc ui
its weird tho how the gradio one slows them down then
a bit, that's gradio stuff a bit slower
they were like slightly slowed down and pitched down
i should still have the samples lemme chekc
oh yes i also sent the resemble enhance result
so u think its a gradio skill issue?
maybe theres smt in the gradio ui code that is wrong
weird
not sure tbh. Needs to be tested

i will prolly check the code of the gradio demo one day
well, about working, it does work, its on CPU, so its completely normal it will take alot of time, id suggest to duplicate it and look at the logs to see the progress, and also to duplicate it on GPU
audiosr has already a gradio ui, its only problem is that its slightly slowed down and pitched down results compared to the no ui inference
how do i duplicate it?
theres no option to duplicate
and how do i get better results????
as of rn, the gradio ui gives slowed down results, id suggest you to use the no ui version until i or it gets fixed as this is an abandoned project, also im working on a no ui colab rn
is there a gpu version of audiosr?
i actually just fixed the slowed results and now the hf space runs on zerogpu
can it run long audio?
how much long
like 3 minuites of audio
mm not really sure, as zerogpu offers max 5 mins of inferencing per session, soo you could test, and if it doesn't work, id just reccomend splitting the audio in more lil ones
how fast is it
Is the huggingface space fixed now?
well that depends by the parameters u set and audio lenght, but its fast
the huggingface space is already fixed, it runs on zerogpu, its ready to use
i got a error how do i fix it
Ayo? @viscid dove level 16 !!! 
What was the error?
If it was GPU task aborted, it was because the file was too long and it's a limitation of zerogpu, which can be fixed by just splitting it
Ifit was about zerogpu quota, it's because there's a quota of free zerogpu per account, it recharges per time
@random ice Can you make a GUI version of audiosr?
Wdym with GUI version as there's already a Gradio UI one ?
For the gpu users
i have a nividia 3080 gpu